Blog for internet marketing


Case Study: Contextual Advertising of a Premium Segment Online Electronics Store: ROI 251%

Service: Google AdWords Contextual Advertising (ads in Google display network)
Project themes: premium segment online electronics store
Region: Ukraine
Total monthly promotion budget: 207,68 USD (including agency's commission)
Promotion period: 12/03/2014 – 28/02/2015

My great delusion: display networking (DN) doesn’t sell. For quite a while, I’ve been thinking that contextual media network is branding, assisted conversions, merged budgets, but not direct selling. At least, not selling something more or less expensive. Once, I've decided to overcome this delusion and started experimenting.

What we had:

  1. A premium segment online electronics store.
  2. Advertising campaigns in display networking for branding.

What we received: 1. Between March 12th, 2014 (the date of launching contextual media network campaigns) to February 28th, 2015 we received 17 transactions for 1899,41 USD, with traffic of 111 490 sessions, and expenses of 4 994.29 USD.

As you can see, the transactions coefficient is vanishingly small. Considering the fact that these stats covers a period of less than one year, the income is also insignificant (to be precise, it averages about 170 USD per month). what we received stats 2. On the other hand, the “%product name% in %store name%” type constructions started to appear in the requests, as well as some branded organic traffic with good transactions coefficient. The branding brought result. 3. In September, we received 69 iPhone 6 pre-orders (conversion from the current preorders is unknown). The cost of preorder was 7 USD. iPhone 6 pre-orders stats

Attempt #1: How to work with display network in a completely wrong way.

We’ve launched the iPhone 6 CMN campaign with two ad groups: the first one with interests targeting, and the second one – with keywords. Here’s what our banners looked like: iPhone 6 frst banners One month passed. We’ve received 1315 clicks, spent 24 USD and got 0 sales.

Conclusion #1.Using a single targeting type doesn’t work. Don’t do that.

Attempt #2: Locating on the platforms

We've analyzed traffic, collected the qualitative platforms (we've selected them by the low bounce rate) and launched one more platform targeted group. In addition, we've manually collected target (in our subjective opinion) platforms. The groups with interest and keyword targeting remained enabled, lowering rates for collecting additional platforms. We've also eliminated about 3,000 rubbish websites with online games for girls. In the end, we've received remarkably little low-quality traffic.

Conclusion #2. The approach to the display network optimization with eliminating low-quality platforms and expecting quality traffic is wrong. There are a plenty of low-quality CMN platforms.

Conclusion #2.1. Topical platform and target audience are not the same thing.

Attempt #3. Learning to combine.

We've analysed the traffic again. We've launched the ad group with the combination of good keywords and good interests (from our point of view). We've stopped all other groups of ad campaigns. Results (the bounce rate and other behavioral factors) saw an improvement, though not a fundamental one. One the current stage, the banners have been replaced, which has also improved CTR and the traffic quality: iPhone 6 ads second type of banners

Conclusion #3. I’ve previously thought that the keywords in contextual media networking function in a very very broad match. “iPhone” and “Buy iPhone” keywords work the same way because in both cases the “iPhone” keyword will appear on the page. To my shame, it appeared to be anything like that. It is necessary to seek for some type of balance between more coverage and more target keywords.

Attempt #4: Studying Google Analytics

After performing the current actions, we've again analysed our campaign trying to understand what had gone wrong. We looked through the “Category interests” report, and saw audiences with good transactions coefficient. We have used them to replace the ones that have been previously selected in accordance with our tastes. The traffic quality improved again, but still there were no sales and the money continued to go wasted.

Conclusion #4. I had previously thought that all audiences in Google Analytics/AdWords completely cover all audiences of CMN. That’s not true. I've checked it within the other project: after adding the whole audience to the remarketing campaign, the amount of impressions has decreased drastically.

Attempt #5: Learning how to write CTA

The traffic quality was already quite good (as we thought) after the previous attempt, and this time we blamed it against the quality of the advertisements themselves. we've provided a small change. The ads started to look like this: iPhone 6 ads third type of banners As a result, CTR improved and traffic quality got extremely close to the quality of traffic from the search. But there still were no sales.

Conclusion #5. I’ve previously thought that if users click on the banner, they are therefore interested (if it’s not an accidental click). That’s not true, the banner quality seriously impacts the traffic quality.

Attempt #6. Delusions are forever.

Once again, we've started an analysing campaign. We have found the report “segments of the audience that is present on the market” (Interested customers audiences (ROI) in AdWords). We've made a segment for users that had bought an iPhone: Interested customers audiences (ROI) in AdWords Audiences with 3% conversion while the average site conversion was 1%?! Here, I understood that the victory was in sight. We've changed the audiences from the “good” ones as in the first report, to “good” ones as in the last one. Together with the improved conversion, the traffic quality has already became better than the one form a search: I’ve been sitting down and waiting for the sales from CMN, nervously rubbing my hands. ads report with good audiences changed

Conclusion #6. The entire list of audiences is not equal to the whole display network audience. Therefore, the segments of the audiences are also not equal in its coverage to the segments, present on the market. People can be interested in the electronics novelties, but don’t plan to buy them.

Conclusion #6.1. Within the last report, I’ve found, in particular, the audience interested in BMW cars. And this audience has given high-quality traffic. If sorting out audiences based on proposals and logics I would have never revealed that feature.

The Happy End

After making a couple of small adjustments (increased rates for several keywords, cutting off several audiences), and three months of suffering, the sales have broken out! Result: result

Expenditure: 207,68 USD
Income: 6407,90 USD
Information from client: 42% of orders made by phone, 30% order, but don’t pay, average margin—11,4%. According to the current data, the figures are as follows:
Profit: 730,50 USD
ROI=((Income*Margin) – Clients’ Expenses) / Clients’ Expenses=((730,50 USD*11.4%) - 207,68 USD) / 207,68 USD = 251%

However, such calculations are not quite correct due to the small number of transactions. We’ve received ROI 191% only from the E-commerce income. As you can see, delusions can ruin your business, and the presented above analysis of contextual advertising case study perfectly demonstrates how it all goes. Display networking optimization is a complex process that requires high-quality content and clearly defined and properly segmented target audience. We have multiply analyzed the campaign itself, used Google Analytics to identify the weak spots, tried different types of banners, etc.

As a result, we have seriously improved sales rate through establishing a strong connection with the target audience. In case you have implemented any of the aforementioned techniques, share your personal experience with us in the commentaries.

Comments (0)

To leave a comment, you have to log in.


to the most useful newsletter on internet marketing


discussed popular read